If Mitt Romney wins tonight, it’ll likely be because of something revealed by a little-noticed statistic released yesterday by the polling firm Rasmussen — following a similar statistic last week from Gallup.
Rasmussen revealed that for the month of October, its data showed that among likely voters, the electorate is 39 percent Republican and 33 percent Democratic.
This comes from a survey of 15,000 people taken over the course of a month. Yes, 15,000 people —15 times the number in a statistically significant poll.
This number might be discounted, since Rasmussen has a reputation as leaning Republican. Except that last week, Gallup — the oldest and most reputable national pollster — released its party ID survey of 9,424 likely voters. And it came out 36 percent Republican, 35 percent Democrat.
This means two firms, with huge numbers of likely voters in their surveys, are finding on election eve that there are more Republican than Democratic voters.
Why does this matter? Because never in the history of polling, dating back to 1936, have self-identified Republicans outnumbered Democrats on Election Day. Never. Ever.
In the two best years for Republicans on this measure, 2004 and 2010, all they could come up with was an evenly divided electorate (37-37 in 2004, down to 35-35 in 2010, according to exit polls).
Yet here we have Rasmussen saying Republicans outnumber Democrats by 6 points and Gallup saying the GOP outnumbers its rival by 1 point.
By some calculations, using 2008 as a benchmark, if the electorate is even or as much as 2 percentage points more Democratic, Mitt Romney will win the popular vote. The closer he gets to a Republican advantage, the more likely it is he wins the Electoral College as well.
Why haven’t we heard more about these party-ID numbers? Pollsters are themselves skeptical of their value in their key role — taking a snapshot of the electorate at a given moment in time.
These numbers are raw data, and so different from polls, which are often “weighted” to reflect demographics. Pollsters take their raw numbers and play around with them to reflect the number of whites, blacks, Latinos, women and the like in the larger population.
They also develop theories, based on past experience and other data, of what the overall electorate is going to look like — then apply those theories to the data as well.
They do not, as a general rule, use party ID as one of these factors, because it is deemed too fluid. Your race and gender don’t change, but you can choose to change your political affiliation at will.
One day someone may call himself an independent, the next day a Republican. A weak Republican may decide to call himself a Democrat. In a right-leaning state like North Carolina, for various reasons, Democrats outnumber Republicans by 20 points.
But presumably, on the eve of an election, when you ask people if they’re Democrats or Republicans or independents, any real movement is going to be between one of the parties and the independent category. There aren’t going to be many people who call themselves Democrats on Oct. 30 and Republicans on Nov. 6.
If this is right — and we’ll get a sense of it today — then these two surveys provide the best backing for the argument that many polls have simply misunderstood the 2012 electorate and weighted their results incorrectly.
One critical feature of the polling, particularly in the key swing states, is that it tends to show a significant margin for Democrats. Polls showing the president ahead by a few points not surprisingly also feature more Democrats — by somewhere between 6 percent and 9 percent.
In 2008, when Obama won Ohio by four points, the Democratic advantage was five points. Move the Democratic advantage down, and you see how the results can alter themselves once people vote.
This use of larger Democratic samples led to objections that Republicans are being systematically or chronically underrepresented, producing a false picture of the electorate.
Note that in certain national polls, Rasmussen’s and Gallup’s among them, Romney led Obama yesterday, 49-48. Neither survey uses its pro-GOP party-ID number as a “weight,” but they’re pretty close to an even electorate. Meanwhile, yesterday’s ABC-Washington Post poll had the president up 50-47 — with a national electorate 6 percent more Democratic than Republican.
If that electorate shows up, Obama wins.
But Gallup and Rasmussen suggest that’s not the electorate, and even a mild representation today of the partisan split their huge surveys found means victory for Romney.
Who wins?
Who knows?